Skip to main content

Risk-averse Multi-stage Mixed-integer Stochastic Programming Problems, İrfan Mahmutoğulları

Tarih: 

Başlangıç Zamanı:  Starting Time 04:00 pm ~ 04:50 pm

Konum:  Rektörlük Toplantı Salonu

Title: Risk-averse Multi-stage Mixed-integer Stochastic Programming Problems

Abstract: In real life, parameters of many optimization problems are not exactly known when the decisions are made. Stochastic programming is an approach for modeling optimization problems that involve uncertainty. In stochastic programming problems, it is assumed that the parameters are not known exactly, but they are distributed with respect to given probability distribution functions.

In this seminar, we present a class of stochastic programming problems where we consider optimal decisions of a risk-averse decision maker in a multi-stage decision environment. We first visit the theory of risk-aversion and models for risk-averse multi-stage mixed-integer stochastic programming problems. Then, we present a scenario decomposition method to obtain bounds on the optimal value of these problems. Using the obtained bounds, we also propose an evaluate-and-cut based exact solution algorithm for the problems with binary first stage variables. We also present the results of computational experiments on the risk-averse multi-stage lot sizing and server location problems. Finally, we consider an application of risk-averse multi-stage mixed-integer stochastic programming in a power system optimization problem.  

 

Bio of the speaker: Ali İrfan Mahmutoğulları is a Ph.D. candidate in Department of Industrial Engineering, Bilkent University.  His research focuses on developing efficient solution methods for risk-averse multi-stage mixed-integer stochastic programming models. He is also interested in application of these models to the problems emerging from different areas of operations research. He is currently doing research as a visiting scholar in H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology.